Detect image masks in TextPage output

Author: JorjMcKie

src/__init__.py

@@ -12585,6 +12585,11 @@ def extractIMGINFO(self, hashes=0):
                 continue
             img = block.i_image()
             img_size = 0
+            mask = img.mask()
+            if mask.m_internal:
+                has_mask = True
+            else:
+                has_mask = False
             compr_buff = mupdf.fz_compressed_image_buffer(img)
             if compr_buff.m_internal:
                 img_size = compr_buff.fz_compressed_buffer_size()
@@ -12600,19 +12605,20 @@ def extractIMGINFO(self, hashes=0):
                     img_size = img.w() * img.h() * img.n()
             cs = mupdf.FzColorspace(mupdf.ll_fz_keep_colorspace(img.m_internal.colorspace))
             block_dict = dict()
-            block_dict[ dictkey_number] = block_n
-            block_dict[ dictkey_bbox] = JM_py_from_rect(block.m_internal.bbox)
-            block_dict[ dictkey_matrix] = JM_py_from_matrix(block.i_transform())
-            block_dict[ dictkey_width] = img.w()
-            block_dict[ dictkey_height] = img.h()
-            block_dict[ dictkey_colorspace] = mupdf.fz_colorspace_n(cs)
-            block_dict[ dictkey_cs_name] = mupdf.fz_colorspace_name(cs)
-            block_dict[ dictkey_xres] = img.xres()
-            block_dict[ dictkey_yres] = img.yres()
-            block_dict[ dictkey_bpc] = img.bpc()
-            block_dict[ dictkey_size] = img_size
+            block_dict[dictkey_number] = block_n
+            block_dict[dictkey_bbox] = JM_py_from_rect(block.m_internal.bbox)
+            block_dict[dictkey_matrix] = JM_py_from_matrix(block.i_transform())
+            block_dict[dictkey_width] = img.w()
+            block_dict[dictkey_height] = img.h()
+            block_dict[dictkey_colorspace] = mupdf.fz_colorspace_n(cs)
+            block_dict[dictkey_cs_name] = mupdf.fz_colorspace_name(cs)
+            block_dict[dictkey_xres] = img.xres()
+            block_dict[dictkey_yres] = img.yres()
+            block_dict[dictkey_bpc] = img.bpc()
+            block_dict[dictkey_size] = img_size
             if hashes:
-                block_dict[ "digest"] = digest
+                block_dict["digest"] = digest
+            block_dict["has-mask"] = has_mask
             rc.append(block_dict)
         return rc
 
@@ -16514,6 +16520,13 @@ def _make_image_dict(img, img_dict):
 def JM_make_image_block(block, block_dict):
     img = block.i_image()
     _make_image_dict(img, block_dict)
+    # if the image has a mask, store it as a PNG buffer
+    mask = img.mask()
+    if mask.m_internal:
+        buff = mask.fz_new_buffer_from_image_as_png(mupdf.FzColorParams(mupdf.fz_default_color_params))
+        block_dict["mask"] = buff.fz_buffer_extract()
+    else:
+        block_dict["mask"] = None
     block_dict[dictkey_matrix] = JM_py_from_matrix(block.i_transform())
 
 

src/extra.i

@@ -3504,10 +3504,11 @@ void JM_make_image_block(fz_stext_block *block, PyObject *block_dict)
 {
     fz_context* ctx = mupdf::internal_context_get();
     fz_image *image = block->u.i.image;
-    fz_buffer *buf = NULL, *freebuf = NULL;
+    fz_buffer *buf = NULL, *freebuf = NULL, *mask_buf = NULL;
     fz_compressed_buffer *buffer = fz_compressed_image_buffer(ctx, image);
     fz_var(buf);
     fz_var(freebuf);
+    fz_var(mask_buf);
     int n = fz_colorspace_n(ctx, image->colorspace);
     int w = image->w;
     int h = image->h;
@@ -3521,6 +3522,8 @@ void JM_make_image_block(fz_stext_block *block, PyObject *block_dict)
         type = FZ_IMAGE_UNKNOWN;
     PyObject *bytes = NULL;
     fz_var(bytes);
+    PyObject *mask_bytes = NULL;
+    fz_var(mask_bytes);
     fz_try(ctx) {
         if (!buffer || type == FZ_IMAGE_UNKNOWN)
         {
@@ -3536,6 +3539,12 @@ void JM_make_image_block(fz_stext_block *block, PyObject *block_dict)
             buf = buffer->buffer;
         } 
         bytes = JM_BinFromBuffer(buf);
+        if (image->mask) {
+            mask_buf = fz_new_buffer_from_image_as_png(ctx, image->mask, fz_default_color_params);
+            mask_bytes = JM_BinFromBuffer(mask_buf);
+        } else {
+            mask_bytes = Py_BuildValue("s", NULL);
+        }
     }
     fz_always(ctx) {
         if (!bytes)
@@ -3559,7 +3568,8 @@ void JM_make_image_block(fz_stext_block *block, PyObject *block_dict)
         DICT_SETITEM_DROP(block_dict, dictkey_size,
                         Py_BuildValue("n", PyBytes_Size(bytes)));
         DICT_SETITEM_DROP(block_dict, dictkey_image, bytes);
-
+        DICT_SETITEMSTR_DROP(block_dict, "mask", mask_bytes);
+        fz_drop_buffer(ctx, mask_buf);
         fz_drop_buffer(ctx, freebuf);
     }
     fz_catch(ctx) {;}

tests/resources/img-regular.pdf

@@ -0,0 +1,31 @@
+"""
+Confirm image mask detection in TextPage extractions. 
+"""
+
+import os
+
+import pymupdf
+
+scriptdir = os.path.abspath(os.path.dirname(__file__))
+filename1 = os.path.join(scriptdir, "resources", "img-regular.pdf")
+filename2 = os.path.join(scriptdir, "resources", "img-transparent.pdf")
+
+
+def test_imagemask1():
+    doc = pymupdf.open(filename1)
+    page = doc[0]
+    blocks = page.get_text("dict")["blocks"]
+    img = blocks[0]
+    assert img["mask"] is None
+    img = page.get_image_info()[0]
+    assert img["has-mask"] is False
+
+
+def test_imagemask2():
+    doc = pymupdf.open(filename2)
+    page = doc[0]
+    blocks = page.get_text("dict")["blocks"]
+    img = blocks[0]
+    assert type(img["mask"]) is bytes
+    img = page.get_image_info()[0]
+    assert img["has-mask"] is True